Home › Intelligent publishing work
Intelligent publishing

AI that need not leave your house.

In open.junixx, AI sits exactly where you work: in the title search, in blurbs and metadata, in written communication and in press analysis. And you decide who does the computing – your own provider, your own key, and a model in your own house if you wish. Then manuscripts, title texts and correspondence stay where they belong.


Artificial intelligence in open.junixx is not a bolted-on extra but a quiet helper – built in at the places where time is lost in a publisher's day.

So the AI search in title data translates your question, in perfectly ordinary language, into a precise query; the AI assistant drafts blurbs and metadata from the existing title data; in Correspondence you rephrase letters and e-mails with a few clicks, or translate them; and in the Media Cockpit the AI recognizes the sentiment of press quotes. The Manuscript review goes furthest: it pre-screens submitted manuscripts and delivers a recommendation, the fit with the lines of your list and editorial notes. The principle stays the same everywhere: the AI makes suggestions – you make the decision.

The AI assistant in correspondence: select the marked text, choose an instruction such as “Phrase more formally” – the AI delivers the matching suggestion, optionally shorter, corrected or translated into another language.
The slot

One connection, three ways.

open.junixx addresses the AI through a single interface. Which model does the computing behind it is a setting – no update, no rebuilding, no change of system. And that choice has a say in whether your texts leave the house at all.

open.junixx one AI interface for every module
  • With a cloud provider

    Anthropic Claude, OpenAI, Hugging Face or any OpenAI-compatible service. Your contract, your key – there is no shared account through which all requests run.

    Requests go to the provider
  • On your own server

    A model on your machine, connected through an OpenAI-compatible address – vLLM, LM Studio or an installation of your own. The same interface, the same templates.

    Requests stay in your network
  • Locally in the house, with no internet at all

    Ollama needs no key and no connection to the outside. Manuscripts, blurbs and correspondence are processed on your own hardware – and only there.

    Not one byte leaves your house

This is set up as master data: Anthropic Claude, OpenAI, Ollama, Hugging Face and any OpenAI-compatible address, Azure, Mistral, Groq, vLLM or LM Studio, for instance. You enter a new model yourself – without waiting for an update from us.

All the brands named on this page belong to their respective owners and are used only to indicate which services open.junixx works with; no partnership or endorsement is implied.

Control

Your AI, your rules.

Who may use the AI, what it may cost and what it may touch is decided by your house. And every intervention stays on the record.

  • Budget with a hard stop

    A monthly budget in euros per tenant. At 80 per cent the system warns; on request it stops hard – and then no request goes out at all.

  • Every request in the log

    Model, token count, cost, duration and status are in the log – including the rejected ones. What the AI has cost is therefore not an estimate.

  • The AI writes only what is allowed

    It may write back only into fields that the template expressly releases. Afterwards the old and the new value are there, with time and person.

  • Keys and rights stay with you

    The access data lie encrypted in your database, no longer readable once saved. Who may use the AI, maintain templates, manage providers, view the log or change the budget is settled by the usual permission management.

FAQ

Frequent questions about the AI interface.

These are the points that come up in almost every conversation – with the answer here.

  • “Does the provider train on our texts?”

    With a model in your own house the question does not arise – there is no provider that receives anything. With a cloud provider it is your contract with them, on their terms: open.junixx is not interposed and has no shared account.

  • “Where is our data?”

    In your database. Only the extract the particular function needs goes to the AI – and every request is in the log with its model, tokens, cost and duration, every field written back with its old and its new value. So you can read afterwards what went out.

  • “And if the provider gets more expensive or shuts down?”

    Then you enter a different one. The model list is master data, the change a setting – no update, no migration, no project. And the cost is capped anyway by a monthly budget with a warning threshold and a hard stop.

Our approach

Assisting, not patronizing.

open.junixx uses AI as quiet support – embedded in your processes, transparent and in your hands.

  • Embedded in the workflow

    AI sits directly in the module concerned – not in a separate tool and not in someone else's browser tab.

  • You stay in control

    The AI makes suggestions; you adopt them deliberately with a click. Nothing is changed unasked.

  • In your own system

    Your publishing data stays in open.junixx – no copying and pasting into external services.

  • Complete without AI too

    AI is an ingredient, not a prerequisite. With no provider set up, open.junixx carries on unchanged – only without the suggestions.

In daily work

Where AI actually helps you.

Five places where open.junixx already takes work off your hands today.

Title search in your own words

Phrase the search like a sentence – open.junixx translates it into a precise database query. No filter building, no learning curve.

  • “All hardcovers by Müller, available” – a hit list straight away
  • Understands ISBN, price, Thema code, year of publication, target group
  • For the whole team – with no filter syntax at all

Blurbs & metadata at the press of a button

From the existing title data the AI assistant generates draft texts – following templates that your house defines itself.

  • Drafts for the blurb, the short description and promotional text
  • Prompt templates maintainable per publisher
  • Uses the title's real data
  • A draft for you to check and polish

Draft letters & e-mails faster

Right in the correspondence editor: select text, choose an instruction, adopt the suggestion.

  • A more formal or a more informal tone
  • Put it more briefly or in more detail
  • Correct the spelling
  • Translate into another language – or your own instruction

Classify press quotes automatically

The Media Cockpit recognizes the sentiment of reviews and makes approved quotes usable for your title data.

  • Automatic sentiment detection (positive, for instance)
  • Approved quotes straight to ONIX
  • Editorial planning across every social channel
  • From the review to the marketing quote

Pre-screen submitted manuscripts

The Manuscript review reads what nobody manages any more: a recommendation, an assessment, the fit with the lines of your list – and notes that turn into editorial work.

  • A quick check on a sample text or a full report on the entire text
  • Fit against your own publishing profiles, exclusion with a reason
  • Tick notes off and adopt them into the work's internal assessment
  • No text goes out without approval in the house and a tick for each run
The benefit

What that means for your publishing house.

Time for what matters

Routine texts and research in seconds instead of hours – more room for work on content.

A consistent brand voice

One tone of voice across title texts, correspondence and social media.

Accessible to everyone

Complex searches and good texts succeed without specialist knowledge or training.

Better metadata & visibility

Complete, meaningful title data – an advantage for the trade, for shops and for discoverability.

Contact

Get to know open.junixx.

Arrange an appointment with no obligation. We will show you how open.junixx fits the way your publishing house works.